Table 6 Attested and unattested units per frequency range
Related Figures (8)
Table 1. Results of primary and secondary RDPs visually. Figure 1 reveals the behavior of each RDP as well as intersections and affinities between semantic categories, where areas of low incidence and areas of high inci- dence can be distinguished. Among the former there are the predicates CAUSE, FROM and usE, where cross-categorial ambiguity is virtually null, and suggests that the NN units under these three RDPs only rarely have a second possible interpretation. These are precise kinds of predicates whose meaning paraphrases are hardly equivocal, e.g. traffic noise can be defined as ‘noise which is caused by traffic’ and is hence assigned the predicate causE. It would seem forced to paraphrase traffic noise through any other ancillary reading, e.g. ‘noise which comes from traffic’ (FROM), or ‘traffic which produces noise’ (MAKE,). The same reasoning applies to (2)-(4): Table 2. Macro-region BE-MAKE-HAVE Table 3. Macro-region ABOUT-FOR-IN Table 4. Corpus frequencies for nouns in modifier and head position Table 5. Productivity models for nouns in modifier and head position malfunction. The results can be interpreted in at least two ways: by focusing on modifier versus head values, the outcome is almost of total balance, with the head values dominating in V/N and in P (3 nouns to 1), and the modifier values being dominant in P* (3 nouns to 1). The second option pays attention to the results for each noun individually, in which case the head position turns out to be more productive for all four nouns (and also for the totals in the last row). Remark- ably, the modifier position appears as most productive according to P*, in theory the most robust formula, and the one which is entirely based on hapaxes. The interpretation must be a prudent one owing to corpus size (2 million words is deemed small for Baayen’s computations), but makes it clear that schema-based measurements are feasible and offer a new path to the study of productivity in compounding (see Baayen, 2010; Maguire et al., 2010). Productivity has been assessed also through lexicographical tools, although such results must be relativized due to the verv nature and purposes of dictionarv
